Activity
From 04/06/2010 to 05/05/2010
05/05/2010
- 09:39 PM Linux kernel client Bug #78: bdi_init list bug
- i suspect this was fixed by commit:5dfc589a8467470226feccdc50f1b32713318e7b
- 09:38 PM Linux kernel client Cleanup #79 (Closed): use bdi setup and register helper
- no. i did rename the bdi ceph-%d, though.
- 09:37 PM Linux kernel client Bug #104 (Resolved): bdi problem on EPERM from osd
- 09:36 PM Linux kernel client Bug #104: bdi problem on EPERM from osd
- problem was use of invalid wbc in completion. fixed by commit:54ad023ba8108d0163acc931ed4b5e4a8a3a7327
- 04:53 PM Linux kernel client Bug #104: bdi problem on EPERM from osd
- the problem is when writepages_finish gets a -1 result code.
- 03:03 PM Linux kernel client Bug #104 (In Progress): bdi problem on EPERM from osd
- 02:57 PM Linux kernel client Bug #104: bdi problem on EPERM from osd
- ladder0:
mount -a
echo asdf > /c/foo
sync
<crash> - 09:04 PM Revision ec600db8 (ceph): osd: fix critical cap parsing bugs and documentation.
- 09:04 PM Revision 642be312 (ceph): rgw: admin uses --secret instead of --key (as --key is already taken)
- 03:56 PM Linux kernel client Bug #107: lockup on __cap_is_valid (via aio_write) vs __ceph_remove_cap
- hmm, s_cap_lock usage sites look okay... don't think it's a leaked spinlock
- 03:03 PM Linux kernel client Bug #107 (Resolved): lockup on __cap_is_valid (via aio_write) vs __ceph_remove_cap
- on master. rsync workload.
[168476.538425] BUG: soft lockup - CPU#0 stuck for 61s! [kswapd0:318]
[168476.538430] ... - 11:45 AM Linux kernel client Bug #106 (Resolved): msgpool depletion?
- which pool is it?
[104608.030333] ceph: msgpool_get ffff88010f1fa370 now 0/0, may fail
[104608.036614] ----------... - 11:40 AM Feature #105 (Resolved): mon: warn on clock drift
- >2s clock drift takes down non-master monitors.. bleh.
- 12:58 AM Bug #98: reserved identifier violation
- I suggest to append a kind of "UUID":http://en.wikipedia.org/wiki/Universally_Unique_Identifier as a suffix to make i...
- 12:37 AM Bug #99: Check return codes everywhere
- Would you like to reuse any class library?
I do not like "assert" for consistent error handling because "the check...
05/04/2010
- 08:57 PM Revision 33b5e426 (ceph): msgr: fix local_pipe peer_type
- Need to re-init_local_pipe() after register_entity().
- 07:57 PM Revision caa77126 (ceph): Merge branch 'mds_snapids' into unstable
- 07:51 PM Revision 89d1b5d0 (ceph): msgr: existing connection may be OPEN on connection race
- Yup!
- 06:56 PM Revision 9bbb545d (ceph): osd: add CMPXATTR defines
- 06:33 PM Revision 3d7d7d1f (ceph): osdmaptool: add -h; don't assert on bad filename
- 03:46 PM Linux kernel client Bug #104 (Resolved): bdi problem on EPERM from osd
- last sysfs file: /sys/class/net/lo/operstate
CPU 1
Modules linked in: ceph [last unloaded: ceph]
Pid: 5724, comm:... - 03:32 PM Bug #99: Check return codes everywhere
- For something like strdup, which fails with ENOMEM, we can throw the usual C++ out of memory exception.
In general... - 01:45 PM Bug #99: Check return codes everywhere
- I suggest to avoid unchecked function calls.
Would you like to detect every error situation as early as possible?
... - 10:56 AM Bug #99: Check return codes everywhere
- Yeah, there are lots of these. Some hosts seem to spit out nice warnings for lots of call sites where the return val...
- 02:08 AM Bug #99 (Closed): Check return codes everywhere
- This issue corresponds to my previous ""feature request":https://sourceforge.net/tracker/?func=detail&atid=763578&aid...
- 02:20 PM Bug #98: reserved identifier violation
- No need for the trailing gibberish, I think.. the trailing _H and preceding CEPH_ should be sufficient.
- 01:38 PM Bug #98: reserved identifier violation
- Leading underscores can be avoided for various identifiers.
How do you think about the application of a name patte... - 10:53 AM Bug #98: reserved identifier violation
- Yeah, I think all the headers #ifdef guards can be changed from __FOO_H to CEPH_FOO_H or similar. Patches welcome!
... - 01:43 AM Bug #98 (Resolved): reserved identifier violation
- This issue corresponds to my previous ""feature request":https://sourceforge.net/tracker/?func=detail&atid=763578&aid...
- 01:21 PM CephFS Bug #52 (In Progress): mds: dentry versionlock
- 01:15 PM CephFS Bug #44 (Closed): Make MDS use new connection-based send_message
- 01:06 PM Bug #62: readdir out of sync on different nodes
- dup #75
- 01:05 PM Bug #62 (Closed): readdir out of sync on different nodes
- 01:02 PM Bug #74: make removed_snaps contiguous
- commit:1c098c23cdb40513c63a723f26f0d416de696f05
- 12:58 PM Bug #74: make removed_snaps contiguous
- 12:57 PM Feature #71: msgr: throttle incoming messages
- 11:32 AM Linux kernel client Bug #28: gracefully fail on fill_trace errors
- this includes ENOMEM on xattr blob
- 10:57 AM Cleanup #100: #define ? enum?
- Sure, these could be changed, but I think it's a pretty low priority item.
- 02:36 AM Cleanup #100 (Closed): #define ? enum?
- This issue corresponds to my previous ""feature request":https://sourceforge.net/tracker/?func=detail&atid=763578&aid...
- 04:22 AM Cleanup #103 (Closed): Introduction of namespaces
- This issue corresponds to my previous "feature request":https://sourceforge.net/tracker/?func=detail&atid=763578&aid=...
- 04:06 AM Revision 32784666 (ceph): ceph.spec.in: run configure
- Reported by MarkN in #ceph on irc.oftc.net
- 04:02 AM Revision 22e9777c (ceph): Merge remote branch 'origin/msgr_throttle' into unstable
- 03:34 AM Bug #102 (Closed): Fix signal handler
- This issue corresponds to my previous ""feature request":https://sourceforge.net/tracker/?func=detail&atid=763578&aid...
- 03:02 AM Feature #101 (Rejected): Conversion of pointer parameters into references
- This issue corresponds to my previous "feature request":https://sourceforge.net/tracker/?func=detail&atid=763578&aid=...
05/03/2010
- 11:57 PM Revision 8a37ac6f (ceph): rbd: update rbd image header format
- 11:57 PM Revision f28decef (ceph): poolop: do op immediately, instead of waiting for timeout
- 11:57 PM Revision cea221c6 (ceph): MPoolOp: rearrange parsing, for easier kernel implementation
- 11:41 PM Revision 1c098c23 (ceph): osd: add removed snapid to removed_snaps for unmanaged snaps
- 11:41 PM Revision 7ec68dcb (ceph): mds: add removal snap_seq to removed_snaps
- This prevents lots of fragmentation in removed_snaps interval set.
- 11:20 PM Cleanup #67: add 'autoscan' items to configure.ac
- This issue fits to my previous feature request ""Complete build configuration template":https://sourceforge.net/track...
- 10:10 PM Linux kernel client Bug #54 (Resolved): do dentry offset assignment when dentry becomes non-null
- 10:09 PM Linux kernel client Bug #54: do dentry offset assignment when dentry becomes non-null
- added to unstable.
- 10:04 PM Revision 14dd8a9a (ceph): mds: acquire anchor_create locks in caller
- We're not allowed to call acquire_locks again once we've completed locking.
Do it in the caller. Add a helper that w... - 09:32 PM Feature #71: msgr: throttle incoming messages
- 09:24 PM Feature #97 (Resolved): qa: better snap tests
- 09:23 PM Feature #96 (Resolved): msgr: close idle connections?
- 09:22 PM Feature #95 (Resolved): mon: adjust overload based on osd disk utilization
- 09:22 PM Feature #94 (Resolved): mon: monitor some osds
- 09:22 PM Feature #93 (Closed): osd: how to tell osd to cleanly shut down, via monitor
- 09:21 PM CephFS Bug #92 (Resolved): mds: journal client_map for rename importing inode
- 09:20 PM CephFS Feature #91 (Resolved): mds: up:shadow mode
- replay client while in standby, so we can take over immediately on failure.
- 09:19 PM CephFS Bug #90 (New): mds: don't sync log on every clientreplay request
- 09:19 PM CephFS Cleanup #89 (Closed): mds: put inode dirty fields in dirty_bits_t to reduce memory footprint
- Most inodes in the cache aren't dirty.
- 09:18 PM CephFS Feature #88 (Resolved): mds: change stray commit strategy to avoid rolling stray dir commits
- Once we commit a temporary item to the stray dir, we have to re-commit later (after purge) to make it go away, at whi...
- 09:17 PM Feature #87 (Closed): osd: separate interface/messenger for backend (osd to osd) network
- 09:16 PM CephFS Feature #86 (Resolved): mds: implement fsck
- rebuild hierarchy after subset of pg's is lost.
and/or, traverse hierarchy and check for errors. - 09:15 PM RADOS Feature #85 (Resolved): osd: pg_num shrink
- 09:15 PM RADOS Feature #84 (In Progress): mon: auto adjust pg_num as pool grows
- of course, splitting needs to be more efficient first!
- 09:14 PM CephFS Feature #83 (New): mds: rename over old files should flush data or revert to old contents?
- write to foo.conf.tmp
close
rename foo.conf.tmp to foo.conf
<crash before flushing new file content>
foo.conf n... - 09:12 PM Feature #82 (Resolved): mon: osd failure smarts
- We shouldn't necessarily believe every osd failure report we get, especially when there are lots of other nodes monit...
- 09:11 PM Revision 53a5b4a5 (ceph): mdsmap: cleanup up braces
- 09:11 PM Revision 11184b90 (ceph): objecter: fix ~Message() calls (use ->put())
- 09:11 PM CephFS Feature #81 (Resolved): mds: do authentication checks
- uid/gid vs file modes, etc.
- 09:09 PM Feature #80 (Resolved): uclient: readdir from cache
- 09:02 PM Revision 409a54f1 (ceph): mds: fix lock->sync rdlock rule
- Observed symptom was touch foo on one node, and missing from ls on another
node.
Since auth rdlock isn't allowed for... - 09:02 PM Revision 632b24f7 (ceph): mds: fix EXCL rdlock rule
- Only the exclusive client should be able to rdlock. (if anyone!)
No observed problems, caught by inspection. - 05:17 PM Revision 291c0ab7 (ceph): filejournal: create if mkfs and journal is small OR dne.
- 04:13 PM Linux kernel client Cleanup #79 (Closed): use bdi setup and register helper
- See commit:e6d086d83cf7f102d48c006f58172a69ec0c15a4
This will make our /sys/kernel/debug/bdi directory pretty (cep... - 04:10 PM Linux kernel client Bug #78 (Resolved): bdi_init list bug
- There were 2 clients mounted, here, so unclear what what was. One was behaving fine.
The other was forcefully unm... - 03:19 PM CephFS Bug #77 (Resolved): mds: anchor_create calls acquire_locks
- fixed in commit:14dd8a9a8a653ef0142ce795454ae3e0e920d386
- 12:57 PM CephFS Bug #77 (Resolved): mds: anchor_create calls acquire_locks
- Locks need to be taken all at once, by caller. Maybe add a helper?
- 02:03 PM CephFS Bug #75 (Resolved): mds: fix rdlock semantics
- fixed by commit:409a54f17515a417f72fb0e1d70733624fe4b609
- 01:57 PM CephFS Bug #75: mds: fix rdlock semantics
- broken result: rdlock succeedes during lock->sync:
10.05.03 13:56:19.316016 7fb391bb0910 mds0.locker rdlock_start... - 01:49 PM CephFS Bug #75: mds: fix rdlock semantics
- rdlock does, actually. The problem was specifically the LOCK->SYNC transition allows rdlock on auth, when it shouldn't.
- 10:09 AM CephFS Bug #75 (Resolved): mds: fix rdlock semantics
- The rdlock does not take which client is doing the locking into account.
Consequently,
clienta$ ls
clientb$ t... - 12:11 PM CephFS Bug #34 (Closed): mds: nonempty cap xlist on snaprealm during trim
- 12:11 PM Bug #48 (Closed): mds: unnecessary snaprealm splits on unlink/rename
- 12:11 PM Bug #53 (Closed): osd: weird crash in snaptrimmer
- 12:11 PM CephFS Bug #51 (Closed): delete directories don't appear in snap
- 12:10 PM Linux kernel client Bug #47 (Closed): gfp at ceph_update_snap_trace+0x16a/0x419
- 12:10 PM Linux kernel client Bug #38 (Closed): rm -r failure
- 12:10 PM Linux kernel client Bug #22 (Closed): BUG at fs/ceph/caps.c:253
- 12:10 PM Linux kernel client Bug #4 (Closed): lockdep warning in socket code
- 12:10 PM Linux kernel client Bug #3 (Closed): leaked dentry ref on umount
- 12:10 PM Linux kernel client Bug #2 (Closed): BUG at fs/ceph/caps.c:2178
- 12:09 PM Linux kernel client Bug #1 (Closed): gpf in tcp_sendpage
- 12:09 PM Bug #68 (Closed): mds/MDCache.cc:4797: FAILED assert(con)
- 12:08 PM Bug #70 (Closed): Interrupting rados tool crashes osd?
- 11:13 AM Linux kernel client Bug #65: crash in tcp_sendpage
- this is probably a problem with the backport.. it went away when we switch to 2.6.34-rc3 on issdm
- 11:09 AM Bug #76 (Resolved): osd: snapdir object recovery doesn't work
- The snapdir object doesn't end up in the missing list.
- 03:50 AM Revision b53d0ed6 (ceph): filejournal: return error instead of asserting
04/30/2010
- 11:24 PM Revision 60095d9d (ceph): mon: correctly update up map when taking over for creating|starting mds
- Otherwise we can get crap like
in <0>
up <{0=4103}>
failed <>
stopped <>
4300: 192.168.141.103:6800/9446 '0' mds0.0 ... - 10:44 PM Revision 3374fbfc (ceph): Merge commit 'v0.20' into unstable
- Conflicts:
src/mon/LogMonitor.cc
src/mon/MDSMonitor.cc
src/mon/Monitor.cc
src/mon/OSDMonitor.cc
src/mon/PGMonito... - 10:11 PM Revision 266d2076 (ceph): debian: fix up debian scripts
- 09:22 PM Revision f244067a (ceph): ceph.spec.in: from Josef
- 09:22 PM Revision b022a285 (ceph): ceph.spec.in: updates
- 08:02 PM Revision aebc239e (ceph): osd: fix pg_to_acting_osds() calculation to consider pg_temp
- The OSD was using pg_to_up_acting, the client uses pg_to_acting, and their
calculations of 'acting' weren't consisten... - 08:02 PM Revision 974f2eee (ceph): mds: fix trim_dentry on dentry under unlinked dir
- We can get a dentry that is trimmable (e.g. null) under a new unlinked dir,
which has no subtree. This will only hap... - 07:58 PM Revision 0b0e1201 (ceph): mds: fix trim_dentry on dentry under unlinked dir
- We can get a dentry that is trimmable (e.g. null) under a new unlinked dir,
which has no subtree. This will only hap... - 07:58 PM Revision 17a2dcbd (ceph): objecter: show osds in laggy debug output
- 05:46 PM Revision 1b3e4fb0 (ceph): osd: fix pg_to_acting_osds() calculation to consider pg_temp
- The OSD was using pg_to_up_acting, the client uses pg_to_acting, and their
calculations of 'acting' weren't consisten... - 05:11 PM Revision 5210d6c3 (ceph): objecter: show osds in laggy debug output
- 03:22 PM Bug #72 (Closed): mon session segfault
- fixed by commit:84c386ea0d255855791fa84d5ab44c6c3a03945a
- 03:21 PM Bug #74 (Resolved): make removed_snaps contiguous
- removed_snaps [2~1,4~1,6~1,8~1,a~1,c~1,e~1,10~1,12~1,14~1,16~1,18~1,1a~1,1c~1,1e~1,20~1,22~1,24~1,26~1,28~1,2...
- 01:03 PM Bug #68 (Resolved): mds/MDCache.cc:4797: FAILED assert(con)
- 11:30 AM Bug #68: mds/MDCache.cc:4797: FAILED assert(con)
- finally found this. commit:dd9bc423a3c95d230772eb3952032f99375973e4
- 04:43 AM Revision 0cb8ad43 (ceph): filestore: fail to mount if journal open fails
- Clean up error reporting too.
- 04:43 AM Revision dc84acd6 (ceph): filestore: fail to mount if journal open fails
- Clean up error reporting too.
- 04:12 AM Revision 7ad42023 (ceph): msgr: fix sigint/term signal handler to actually quit (by chaining)
- 04:12 AM Revision 3480dc69 (ceph): msgr: fix sigint/term signal handler to actually quit (by chaining)
04/29/2010
- 10:05 PM Revision 269d1426 (ceph): librados: fix blank pool stats
- 10:04 PM Revision 96a90efc (ceph): librados: fix blank pool stats
- 09:30 PM Feature #73 (Rejected): kernel: implement 'flush disk cache' ioctl
- Currently the osd journal on a raw block device requires that you disable the disk write cache via hdparm. That's an...
- 09:29 PM Revision 61b4b295 (ceph): qa: make pjd clean up
- 09:29 PM Revision ebe05ae6 (ceph): mds: clean up MIN/MAX, client_ranges
- Avoid double evaluation via macros...
- 09:29 PM Revision d595ad44 (ceph): msgr: remove pid file on SIGINT, SIGTERM
- 09:28 PM Revision 3a9439ca (ceph): mds: clean up MIN/MAX, client_ranges
- Avoid double evaluation via macros...
- 09:28 PM Revision 8fcb0cc4 (ceph): qa: make pjd clean up
- 09:28 PM Revision 30561d84 (ceph): msgr: remove pid file on SIGINT, SIGTERM
- 06:57 PM Revision a31ea46f (ceph): config: post-process default config options too
- Previously we were only doing the ones we read out of the .conf.
- 06:56 PM Revision d1441462 (ceph): config: post-process default config options too
- Previously we were only doing the ones we read out of the .conf.
- 05:43 PM Revision 318dfbeb (ceph): mon: use (robust) buffer::list::write_fd() to write to disk
- 05:43 PM Revision 63385d02 (ceph): mon: use (robust) buffer::list::write_fd() to write to disk
- 05:27 PM Revision 1f55b175 (ceph): mon: m->get_session() may return null if session has closed
- because the session close clears connection->priv. We need to check at
each site anyway, either for null session, or... - 05:22 PM Revision 84c386ea (ceph): mon: m->get_session() may return null if session has closed
- because the session close clears connection->priv. We need to check at
each site anyway, either for null session, or...
04/28/2010
- 10:55 PM Revision a846236b (ceph): configure: remove second AM_INIT_AUTOMAKE line
- 09:22 PM Revision 6e7ff1a2 (ceph): msgr: fix segfault sending msg on closed Connection/pipe
- 09:18 PM Revision 48c086a5 (ceph): string fix
- 09:07 PM Revision 2e2222f8 (ceph): msgr: fix segfault sending msg on closed Connection/pipe
- 08:56 PM Revision 59e77b35 (ceph): msgr: add message throttling; with a default max of 100MB waiting for d...
- 08:45 PM Revision b8cca8f3 (ceph): replace nstring with std::string
- 05:36 PM Revision 07e1a2f6 (ceph): throttle: add fairness
- This assumes that SignalOne() will wake people up in the right order. Not
sure if that's true... - 04:51 PM Bug #72 (Closed): mon session segfault
bt
#0 0x00000000005b65d4 in AO_fetch_and_add_full (p=0x8, incr=18446744073709551615) at /usr/include/atomic_ops/s...- 02:24 PM Bug #70 (Resolved): Interrupting rados tool crashes osd?
- caused by sending message on closed Connection with null pipe.
fixed by commit:6e7ff1a2150618a989d78574a11ddac91d... - 01:36 PM Cleanup #58 (Resolved): kill nstring/cstring, use std::basic_string instead
- commit:ff8abee39f55e4222049145974a37fb395fba6ed
- 10:43 AM Feature #71 (Resolved): msgr: throttle incoming messages
- see msgr_throttle branch. should be able to use common/Throttle.h.
probably need to verify that it is indeed fa... - 10:10 AM Bug #68: mds/MDCache.cc:4797: FAILED assert(con)
- wido hit this too. rsync workload.
04/27/2010
- 05:00 PM Bug #70 (Closed): Interrupting rados tool crashes osd?
- 432 16 2096 2080 19.2568 24 3.14879 3.31748
^C
gregf@pudgy:~/ceph/src$ ==30658==
==3...
04/26/2010
- 09:51 PM Revision 218f031a (ceph): mds: fix snaprealm split on capless inode
- 09:08 PM Revision f688ffab (ceph): mds: fix snaprealm split on capless inode
- 09:08 PM Revision b3bb8c60 (ceph): buffer: clean up namespace pollution
- 08:46 PM Linux kernel client Bug #69 (Can't reproduce): ceph: ffff88001976ba50 auth cap (null) not mds0 ???
- no apparent malfunction. on master branch. dmesg is
ceph2 login: [249643.959209] ceph: get_reply unknown tid 260... - 03:54 PM Bug #68 (Closed): mds/MDCache.cc:4797: FAILED assert(con)
- 10.04.26 22:33:30.354868 428a1940 -- 10.178.28.97:6800/10059 --> client7202 10.178.28.97:0/1984433234 -- client_reply...
- 10:54 AM Cleanup #67 (Closed): add 'autoscan' items to configure.ac
04/24/2010
- 05:10 PM Revision 6ec8870d (ceph): poolop: rename and move definitions to ceph_fs.h
- This was done so that we can use it in the kclient
04/23/2010
- 10:50 PM Revision 0decb33d (ceph): Makefile: define __CEPH__ in ceph builds. use it.
- Avoid breakage when headers are used outside of ceph build (librados-dev)
- 10:22 PM Revision a6ee3aef (ceph): mds: fix req use-after-free to get connection to direct reply
- 10:13 PM Revision 282b2428 (ceph): atomic: reverse if test so we don't break things on separately-compiled...
- 10:12 PM Revision 07dea448 (ceph): mds: remove unused variable from Locker
- 09:50 PM Revision 3ed1a55d (ceph): debian: build-depends libatomicops-dev
- 09:47 PM Revision e5a2080d (ceph): Merge branch 'snaps' into unstable
- 09:46 PM Revision 8f5d5ea7 (ceph): mon: clean up mon snap type checks
- 09:42 PM Revision 4a53f1c6 (ceph): librados: clean up headres
- 09:42 PM Revision ca464284 (ceph): mon: don't create/delete unmanaged snaps if pool snaps exist
- 09:42 PM Revision cf8ce29d (ceph): osd: no error checking in data structure (caller should do it)
- 09:42 PM Revision 503c41e6 (ceph): mon: change the unmanaged snap interface for better distinction of repl...
- and snapid numbers.
Of course, this now means that the snapid isn't actually passed
back to the client, but that is ... - 09:42 PM Revision 57064765 (ceph): msg: Give MPoolOpReply a bufferlist for reply data
- 09:42 PM Revision f612ab86 (ceph): mon: OSDMonitor fills in MPoolOpReply's bufferlist on unmanaged snap cr...
- 09:42 PM Revision 076517fb (ceph): objecter: Implement bufferlist-based return code for pool ops
- 09:42 PM Revision 832c9192 (ceph): librados: Adjust snapid allocation ufnctions and implement new objecter...
- 09:42 PM Revision ec7ecc57 (ceph): objecter: do pool op result parsing in Objecter completion wrapper
- 09:42 PM Revision 0fb4a6e8 (ceph): librados: Set selfmanaged snap functions to use int instead of int64_t,...
- 09:42 PM Revision d1d235e7 (ceph): librados: add functions for client-managed snapshots in all the interfaces
- 09:42 PM Revision 4b0021fb (ceph): mon: don't crash if a client tries to self-manage managed snaps, or vic...
- 09:42 PM Revision 39817b3d (ceph): librados: add small bit of documentation, fix headers
- 09:42 PM Revision 5116700f (ceph): msg: Add space for snapid in MPoolOp. Reply now uses an int64_t replyCode.
- This allows space for a full 64-bit snapid to be returned, which is about
to be necessary. - 09:42 PM Revision ac129d78 (ceph): osd: pg_pool_t gets new functions for unmanaged (ie, client-managed) snaps
- 09:42 PM Revision eba044ce (ceph): mon: OSDMonitor handles client-managed snapshot requests.
- 09:42 PM Revision 5fb4e8fa (ceph): objecter: New functions for client-managed snapshots.
- 09:30 PM Revision 4fda5f92 (ceph): msgr: fix atomic_t test
- Missed this in merge 50285ecf5c404b456a0427eb7d6d1097ffc0ea83
- 09:22 PM Revision 5b6cfff3 (ceph): Merge branch 'atomic' into unstable
- 09:22 PM Revision 50285ecf (ceph): Merge branch 'msgr' into unstable
- Conflicts:
src/TODO
src/mds/Server.cc
src/msg/SimpleMessenger.cc
src/msg/SimpleMessenger.h
src/osd/OSD.cc - 08:22 PM Linux kernel client Bug #66 (Resolved): BUG_ON(req->r_reply) at fs/ceph/mds_client.c:1841!
- [ 6447.063496] ------------[ cut here ]------------
[ 6447.065210] kernel BUG at fs/ceph/mds_client.c:1841!
[ 6447.... - 08:13 PM Feature #60 (Resolved): librados: selfmanaged snaps
- 08:13 PM Feature #21 (Resolved): optionally use libatomic for atomic_t
- 07:10 PM Revision 756b251b (ceph): v0.20
- 06:07 PM Revision 609eaa14 (ceph): mds: skip client snap notification on unlink
- This is cheating a bit, but should be harmless. Basically, we spit off the
snaprealm when we unlink to keep the hier... - 04:13 PM Linux kernel client Bug #65 (Resolved): crash in tcp_sendpage
- on issdm. master branch (standalone).
[ 2900.360800] ceph: osd16 weight 0x10000 (in)
[ 2900.360802] ceph: osd17 ... - 12:28 PM Linux kernel client Bug #64 (Resolved): crash in handle_mds_map (corrupt s_waiting list?)
- unstable branch.
ceph: mds0 caps stale
ceph: mds0 caps stale
ceph: mds0 hung
ceph: mds0 came back
ceph: mds0 c... - 11:08 AM Bug #48 (Resolved): mds: unnecessary snaprealm splits on unlink/rename
- worked around this in commit:609eaa14ca660a9b8b965f5713921ce4be5c80be
- 10:55 AM Linux kernel client Bug #63 (Resolved): dentry_info slab not empty
- [97009.315064] slab error in kmem_cache_destroy(): cache `ceph_dentry_info': Can't free all objects
[97009.324159] P... - 10:49 AM Bug #62 (Closed): readdir out of sync on different nodes
- [09:56:33 eestolan@issdm-7:~] for i in 3 7 8; do ssh issdm-$i "hostname; mount | grep ceph; ls /mnt/ceph; echo"; done...
04/22/2010
- 10:12 PM Revision 4d7b2ceb (ceph): mds: add CompatSet to mdsmap
- 10:12 PM Revision feca69da (ceph): mds: verify mdsmap compatset on startup
- 10:12 PM Revision f502547f (ceph): osd: print a bit more debug info about compat problems
- 10:12 PM Revision 7a8daedf (ceph): osd: more useful debugging on repop version, out of order completion
- 09:40 PM Revision bfd96295 (ceph): rbd: change rbd header disk format, update protocol version
- 08:10 PM Revision e3ef1484 (ceph): msgr: never (re)open pipe when sending message directly to Connection*
- 07:04 PM Revision 76320373 (ceph): mdsmap: add separate struct version for extended portion of mdsmap
- 06:43 PM Revision d03bddc6 (ceph): cauthtool: add --gen-print-key
- Just generate a random cryptokey and print it out.
- 06:39 PM Revision 5d2dcce7 (ceph): auth: take key or file containing key on command line
- e.g.
ceph --key AQBggdBLCJwLABAAuPVK/qJ3/ySbRvEnJHNJ8w== -s
ceph --keyfile file_with_key -s
ceph -K file_with_key -... - 06:27 PM Revision 2acb843b (ceph): cauthtool: always print entire encoded CryptoKey
- Make cauthtool -l output consistent with what you need to feed to
mount.ceph(8). - 02:37 PM Feature #60 (Resolved): librados: selfmanaged snaps
- 02:37 PM Revision 0fc47e26 (ceph): msgr: increment connect_seq when STANDBY->CONNECTING
- Otherwise we get this on the peer:
msg/SimpleMessenger.cc: In function 'int SimpleMessenger::Pipe::accept()':
msg/Si... - 02:24 PM Bug #59 (Resolved): max_size of 1
- client says:
[21764.780480] ceph: file.c:824 : aio_write ffff88010bcdf5d0 10000122416.fffffffffffffffe 0~7... - 01:55 PM CephFS Feature #57 (Resolved): mds: feature bits in mdsmap?
- 09:52 AM Feature #21: optionally use libatomic for atomic_t
- Huh, it looks like STL has some refcounting primitives built in?
From /usr/include/c++/4.3/bits/basic_string.h:
... - 09:46 AM Linux kernel client Feature #42: Resize of rbd image
- rbdtool part is done. i think the driver just needs to have a 'refresh' function to reload the image metadata.
- 09:42 AM Cleanup #58 (Resolved): kill nstring/cstring, use std::basic_string instead
- I created cstring because the stl spec says that c_str() may be a temporary value, i assumed because the string wasn'...
04/21/2010
- 10:28 PM Revision 62985351 (ceph): Merge remote branch 'origin/mdssnaps' into unstable
- 10:23 PM Revision 44208739 (ceph): filestore: default to writeahead journal, and no btrfs snaps
- At least until btrfs snap deletion doesn't require a full commit (i.e. each
commit cycle doesn't do a commit for the ... - 09:56 PM Revision 2caf5135 (ceph): rbdtool: add, use prettybyte_t type
- 09:52 PM Revision 0cf079ac (ceph): rbdtool: add --info, --resize commands
- 09:52 PM Revision f4ef6a93 (ceph): auth: make 'admin' the default user
- We don't create a 'guest' user anyway.
- 09:24 PM Revision d5c89276 (ceph): sample.ceph.conf: add example on writeahead (vs parallel) journal
- 09:24 PM Revision 99ac12b1 (ceph): librados: adjust C++ aio completion creation api
- 09:24 PM Revision f2160f1a (ceph): osd: create 'rbd' pool
- 09:06 PM CephFS Feature #57 (Resolved): mds: feature bits in mdsmap?
- 05:58 PM Revision 8478ebed (ceph): debian: require hdparm, recommend btrfs-tools
- 05:58 PM Revision d4bf8eef (ceph): debian: require hdparm, recommend btrfs-tools
- 03:30 PM Bug #56 (Closed): osd: crash on repop completion
- 10.04.21 13:51:55.335613 7f3d08fd1910 journal throttle: waited for bytes
osd/ReplicatedPG.cc: In function 'void Repl... - 03:28 PM Bug #55 (Resolved): osd: fix transition from snaps -> no snaps -> snaps
- After filestore btrfs snaps = 1 -> 0, we may have leftover snap dirs. Later, if we turn them back on again, we'll ro...
04/20/2010
- 11:39 PM Revision c719f4e4 (ceph): mds: don't cow already cowed dentry
- 11:39 PM Revision 327432ae (ceph): mds: rename move_to_realm(), add debugging (cleanup)
- 11:39 PM Revision 476606f4 (ceph): mds: always use get_newest_seq (not snap) when doing cow
- 11:38 PM Revision 93058af3 (ceph): sample.ceph.conf: updated
- 11:38 PM Revision ba5a0d1f (ceph): fetch_config: enable fetching ceph.conf from a remote location
- fetch_config: always fetch if script exists and is executable
fetch_config: make /etc/ceph/sample.fetch_config +x
M... - 05:47 PM Revision 5a5cc183 (ceph): osd: don't capture SIGINT/SIGTERM; journal and/or btrfs snaps are suffi...
- We used to do this to avoid corrupting the filestore, but since we can now
either roll forward with the writeahead jo... - 05:39 PM Revision f7f6daf1 (ceph): auth: less noisy debug
- 04:58 PM Revision 34d30877 (ceph): mds: drop mutation locks
- Saw this one, but fixed others that appear broken as well:
mds/MDCache.h: In function 'virtual Mutation::~Mutation()... - 01:35 PM Linux kernel client Feature #23: fcntl/flock advisory lock support
- All right, designed a basic interface between the client and the MDS. Going to implement the MDS and messaging parts ...
- 10:07 AM Linux kernel client Bug #54 (Resolved): do dentry offset assignment when dentry becomes non-null
- see rename_vs_i_complete branch on fatty for partial solution
- 03:43 AM Revision d8f1de5e (ceph): osd: fix snapset loading from _snapdir
- Another masked variable! In the same block as the last one! Aie.
- 03:29 AM Revision 1915f2d5 (ceph): mds: clone dentry for multiversion dir if linkage is changing...
- ...even if the inode itself doesn't need to be cowed. In particular, we
do the pre_cow_old_inode() thing, so it freq...
04/19/2010
- 11:38 PM Revision 07c7873d (ceph): mds: avoid null deref when hiding .ceph dir
- 10:37 PM Revision f9225fe6 (ceph): mds: xlock dentry versionlock to cleanly avoid projected dn linkage rep...
- 10:36 PM Revision 463792de (ceph): reorder locks
- 10:36 PM Revision c8d7c970 (ceph): mds: always issue lease mask 1; assume CEPH_LOCK_DN
- This makes it possible for us to change CEPH_LOCK_DN definition without
breaking the client. - 10:35 PM Revision beb562d9 (ceph): mds: add versionlock to dentry
- 09:32 PM Revision 4f402cb4 (ceph): mds: rename send_message_client to send_message_client_counted
- 09:22 PM Revision 09366cbf (ceph): mds: fix erroneous uses of send_message that need to be send_message_cl...
- 09:22 PM Revision 3484b844 (ceph): mds: Add new Connection-based send_message_client that uses inc_push_se...
- 08:43 PM Bug #53 (Resolved): osd: weird crash in snaptrimmer
- fixed in commit:d8f1de5e87114670dca08c9a942d6b1cd39f40a7
another stupid masked variable! int r this time. - 08:34 PM Bug #53 (Closed): osd: weird crash in snaptrimmer
- segfault when printing the snapset ref, even though we printed it a few lines earlier.
pretty reproducible when de... - 08:27 PM CephFS Bug #51 (Resolved): delete directories don't appear in snap
- fixed by commit:1915f2d590fe571fe92e692abd9a51d00169838d
- 04:37 PM CephFS Bug #51 (Closed): delete directories don't appear in snap
- mkdir a
mkdir .snap/one
rmdir a
ls .snap/one # a is not there
- 08:04 PM CephFS Bug #52 (Resolved): mds: dentry versionlock
- add dentry versionlock to avoid pipelining dentry linkage changes across multiple mds's (projected linkages aren't re...
- 03:51 PM Linux kernel client Bug #50 (Resolved): osd timeout reset leaves some ops hanging
- i see an osd reset:
[ 7769.416465] ceph: tid 65961 timed out on osd4, will reset osd
followed by stray replies... - 02:32 PM CephFS Bug #44: Make MDS use new connection-based send_message
- Wrote a new send_message_client that properly calls inc_push_seq and switched the appropriate calls to use it, resolv...
- 10:30 AM Bug #48 (In Progress): mds: unnecessary snaprealm splits on unlink/rename
- 10:26 AM Linux kernel client Feature #23 (In Progress): fcntl/flock advisory lock support
- 10:20 AM Linux kernel client Bug #36 (Rejected): uml crash in sendpage
- can't reproduce. bad compilation?
- 10:19 AM Linux kernel client Bug #47 (Resolved): gfp at ceph_update_snap_trace+0x16a/0x419
- bug in realm split. fixed by commit:be23f8ad6fccba9a5535a7d3013eb492c20e39cb
- 08:59 AM Linux kernel client Bug #47 (In Progress): gfp at ceph_update_snap_trace+0x16a/0x419
- 08:58 AM CephFS Feature #45: Investigate adding Connection * to mds Session
- Take a look at send_message_client(Message *, client_t) and send_message_client(Message *, entity_inst_t). They both...
04/16/2010
- 10:55 PM Revision 65ec0241 (ceph): autoconf: Fix detection of sync_file_range.
- Without this patch, on CentOS 5.4 ./configure reports that
sync_file_range is missing, but HAVE_SYNC_FILE_RANGE ends
... - 10:51 PM Revision a0e7d8df (ceph): remove references to [group] and [mount] in sample.ceph.conf
- 10:18 PM Revision d2af75de (ceph): add executables to .gitignore
- 09:55 PM Revision edf935d1 (ceph): mds: Fix bad pointer dereference brought about by switch to Message ref...
- 08:52 PM Revision d8be246b (ceph): mds: xlock versionlock on rename if witnesses
- This ensures that we don't pipeline dentry linkage updates when there
are witnesses. That can cause problems because... - 07:30 PM Revision 06e5528a (ceph): mds: use Connection-based send_message in handle_discover
- 07:28 PM Revision 6a0d0d06 (ceph): msg: Remove "asker" data member from MDiscover.
- Fix uses.
- 07:28 PM Revision 751a4056 (ceph): mds: Fix some compile-time issues.
- 06:12 PM Revision da13000d (ceph): mds: Subsystems generally use the connection-based send_message when po...
- This isn't a very deep change, though -- a lot of the remaining instances
could probably be fixed up just by changing... - 04:36 PM Revision 05b0e3a5 (ceph): mds: create send_message(Message *m, Connection *c) passthrough
- 04:07 PM Linux kernel client Bug #38 (Resolved): rm -r failure
- the problem is (mostly) that d_move reorders d_subdirs. so just clear I_COMPLETE on rename.
but also:
- we shoul... - 03:59 PM Bug #40 (Closed): remove references to [group] and [mount] conf sections from wiki, sample .conf ...
- 03:49 PM Bug #40 (In Progress): remove references to [group] and [mount] conf sections from wiki, sample ....
- 02:56 PM CephFS Bug #46 (Closed): Remove asker from MDiscover
- Yep, that's done. Got worried by a bug, but it was unrelated (and fixed now!).
- 11:22 AM CephFS Bug #46 (In Progress): Remove asker from MDiscover
- 11:22 AM CephFS Bug #46 (Closed): Remove asker from MDiscover
- It's redundant and confusing with get_orig_source(); and it's unecessary since MDiscover messages aren't forwarded an...
- 01:17 PM Bug #48 (Closed): mds: unnecessary snaprealm splits on unlink/rename
- 12:30 PM Linux kernel client Bug #47 (Closed): gfp at ceph_update_snap_trace+0x16a/0x419
- after running snaptest1.sh twice:
[90471.092551] general protection fault: 0000 [#1] PREEMPT SMP
[90471.095369] ... - 11:21 AM CephFS Feature #45 (Closed): Investigate adding Connection * to mds Session
- This would open up several more uses of connection-based send_message.
- 11:18 AM CephFS Bug #44: Make MDS use new connection-based send_message
- Implemented a send_message(Message *, Connection*) in MDS, and switched a lot of code to use it. I'm going to see if ...
- 12:04 AM Revision b4d42315 (ceph): rados: aio, reorder callback completion handler creation
04/15/2010
- 11:34 PM Revision bd34079e (ceph): mds: set trivial uses of send_message to use the Connection version.
- 10:58 PM Revision f6d26664 (ceph): osd: Send messages via Connection * when possible.
- Also, rename a message shadowing its trigger message so that we can.
- 10:58 PM Revision b38cebf0 (ceph): osd: send reply via connection for MOSDSubOps
- 10:58 PM Revision 69682ad8 (ceph): osd: send_message via Connection when possible
- 10:58 PM Revision 85b67cec (ceph): msgr: Implement new send_message[_lazy](Message *m, Connection *con) fu...
- These can reduce the number of lookups and should be able to recover
from any pipes being destroyed while the Connect... - 09:44 PM Revision 47f5315f (ceph): filestore: be thorough about wiping out old osd data content during mkfs
- Delete any subvols or directories that we fine.
- 08:43 PM Revision 057897e4 (ceph): msgr: if a pipe** is passed in to submit_message, it's used.
- This is a double pointer because it's possible that a Pipe has
been destroyed since the send_message call, so we want... - 08:43 PM Revision 6c9bee49 (ceph): msgr: Add a void *pipe to Connection.
- This is set on Connection creation and zeroed out in ~Pipe().
- 08:43 PM Revision 731414b2 (ceph): msgr: give submit_message new interface.
- 08:43 PM Revision a9800e73 (ceph): msgr: remove unnecessary clone variable from submit_message
- 06:31 PM Revision fd28d9a8 (ceph): cephfs: include LAZYIO caps if LAZY file mode bit is set
- 05:12 PM Revision 06d032e8 (ceph): filestore: log snap_create error before asserting success
- 05:09 PM Revision 2cc0ec29 (ceph): mon: do not build mon to mon authorizer if cephx is disabled
- 04:46 PM CephFS Bug #44 (In Progress): Make MDS use new connection-based send_message
- 04:44 PM CephFS Bug #44: Make MDS use new connection-based send_message
- All the straight-up uses of send_message which are replying to another message have been modified. Still need to add ...
- 04:42 PM CephFS Bug #44 (Closed): Make MDS use new connection-based send_message
- This will be more complicated than the OSD adjustments because there are a lot of send_message_[client|mon|mds|osd] c...
- 04:40 PM Feature #43 (Resolved): Make OSD use new connection-based
- The OSD now uses this method to send_message whenever possible. There are a bunch of if-else blocks to allow for forw...
- 04:39 PM Feature #43 (Resolved): Make OSD use new connection-based
- We have a new send_message(Message *m, Connection *c); let's use it in the OSD!
- 04:37 PM Feature #32 (Resolved): msgr: send messages via a Connection*
- This is done and pushed in the msgr branch. I'm creating separate issues to track completion progress in the MDS, OSD...
- 10:49 AM Bug #39 (Closed): Re-examine PaxosServiceMessage::get_session() and use
- Can't reproduce. The code actually had a bad memory leak at the time and was causing the machine to run out of memory...
- 10:30 AM Linux kernel client Feature #42 (Resolved): Resize of rbd image
- We want to be able to resize the rbd image. Snapshots should be maintained and should have a size associated with them.
- 10:30 AM Linux kernel client Bug #38 (In Progress): rm -r failure
- reliably reproducible with kernel_untar_build.sh workunit. looks like problem with dcache_readdir. from the rm -r's...
- 10:26 AM rbd Feature #41 (Resolved): Support snapshots
- We'd like to be able to support snapshots in both the kernel driver and in the kvm-rbd.
- 09:58 AM Bug #40 (Closed): remove references to [group] and [mount] conf sections from wiki, sample .conf ...
04/14/2010
- 11:54 PM Revision 69b849c4 (ceph): buffer: Include Spinlock.h.
- This is inherited from one implementation of atomic_t but not necessarily all.
Including the file here is necessary i... - 11:54 PM Revision d1d6da3b (ceph): atomic_t: Remove (never used) unstable commonc++-based version
- 11:54 PM Revision ca74b4bb (ceph): atomic_t: Add a new version based on libatomic_ops.
- Hopefully this version will take less CPU since it isn't running Spinlocks, but
we'll need to test for that.
Includes... - 11:16 PM Revision d0213ddd (ceph): mds: allow LAZYIO cap along with WR
- 10:18 PM Revision 198d7106 (ceph): osd: fix source type checks (osd, mon)
- The peer type is on the connection, no need to look at Session. Also, we
don't currently have Sessions for outgoing ... - 10:08 PM Revision afcee3b0 (ceph): osd: don't ->put() a null session in _share_map_incoming
- 09:47 PM Revision ec8471d6 (ceph): osd: queue pg update under pg sequencer
- 09:47 PM Revision 69b108f8 (ceph): osd: queue pg creation under pg sequencer
- Avoids pg log, metadata writes from being reordered. Had to refactor
try_create_pg to be less annoying. - 07:40 PM Revision 09916d35 (ceph): qa: set -e in workunits
- 07:40 PM Revision cfabc020 (ceph): mds: not dout(0) request latency
- 06:31 PM Revision 96befd4a (ceph): mds: fix broken snaprealm split optimization for (unlinked) files
- Splitting a file is trivial; do it. No special case for 'stray'.
- 05:16 PM Bug #39 (In Progress): Re-examine PaxosServiceMessage::get_session() and use
- 04:56 PM Bug #39 (Closed): Re-examine PaxosServiceMessage::get_session() and use
- I got a crash while testing the new libatomic-based atomic_t, but I don't think it could have been caused by the chan...
- 05:09 PM rgw Support #8: Document differences from S3
- I'll probably just do this when I do the panel info for Kyle, when he asks. It doesn't need a due date.
- 05:00 PM Feature #21: optionally use libatomic for atomic_t
- I've cleaned out all the CommonC++ stuff, which couldn't be used anyway since it didn't match the required interface!...
- 03:34 PM Linux kernel client Bug #38 (Resolved): rm -r failure
- seeing this one again on current clients (commit:a6a5349)
- 03:34 PM Bug #17 (Rejected): rm -r failure
- wrong project
- 03:32 PM Bug #37 (Rejected): osd: recover missing clone object
- missing clone object.. it's in the pg log, but not there. crash in push_to_replica during recovery.
- 03:31 PM Linux kernel client Bug #36 (Rejected): uml crash in sendpage
- running kernel untar qa workunit on fatty.
with valgrind on osd - 02:49 PM Bug #35 (Duplicate): osd: pg split should queue transaction(s) under appropriate sequencer(s)
- any pg metadata writes need to be done under the pg sequencer to avoid reordering things like write_info, write_log a...
- 11:56 AM rgw Bug #31 (Closed): Handle mal-formed requests better
- HTTP_HOST is now an optional header, and I didn't notice any others behaving badly.
Plus it helps if your client a... - 11:49 AM CephFS Bug #34 (Resolved): mds: nonempty cap xlist on snaprealm during trim
- should be fixed by commit:96befd4a5ad568fe34faab0c56edbd3a4a92a6c3
- 09:57 AM CephFS Bug #34 (Closed): mds: nonempty cap xlist on snaprealm during trim
- ./include/xlist.h: In function 'xlist<T>::~xlist() [with T = Capability*]':
./include/xlist.h:63: FAILED assert(_siz...
04/13/2010
- 10:48 PM Revision 6573493a (ceph): rgw: don't crash if HTTP_HOST isn't set
- 07:27 PM Revision 69fda2a9 (ceph): osd: fix loading of snapset from _snapdir object attr
- This happens when the head is deleted but we still have snaps. A nested
bufferlist declaration was masking things. - 07:27 PM Revision a6c80a3a (ceph): osd: avoid PGPool lookup in inner loop
- Each PG already has a PGPool *.
- 06:49 PM Revision 50cad814 (ceph): osd: add a message_from_mon function
- 06:49 PM Revision e5cd2b72 (ceph): osd: replace the ALLOW_MESSAGES_FROM macro with use of OSDCaps functions.
- Now with less null pointer exception! Code checks for session
before dereferencing it. - 06:03 PM Revision c94e4904 (ceph): Revert "osd: replace the ALLOW_MESSAGES_FROM macro with use of OSDCaps ...
- This reverts commit 9ceb3be9a62f7b777b23a4d65b1cc302ec834e6b.
This will be coming back in another form shortly, but a... - 05:38 PM Revision 81e748df (ceph): mds: only recover files from stale clients when auth
- Replicas have client_ranges defined too, but only the auth MDS should do
the file size/mtime recovery.
Also rename c... - 04:51 PM Feature #33 (Closed): O_LAZY or equivalent
- - revisit mds/caps infrastructure to support this
- come up with a suitable api (probably ioctls?) for setting lazy ... - 04:03 PM Feature #32 (Resolved): msgr: send messages via a Connection*
- Since we usually have a Connection* handy, add support for sending a Message to a Connection directly and avoid the l...
- 02:09 PM Linux kernel client Bug #4 (Resolved): lockdep warning in socket code
- 02:09 PM Linux kernel client Bug #4: lockdep warning in socket code
- use separate class for ceph socket.
commit:79cb735c - 01:21 PM rgw Bug #31 (Closed): Handle mal-formed requests better
- RGW can't handle requests that don't include things like the HTTP_HOST header -- it just crashes out.
This was enc... - 12:44 PM CephFS Bug #30 (Resolved): multimds: slave_request on getattr
- 10.04.13 12:39:39.111206 408a7950 -- 10.0.1.245:6802/14233 <== client4122 10.0.1.141:0/3071614701 46048 ==== client_r...
- 11:56 AM Feature #10 (Resolved): osd: Replace ALLOW_MESSAGES_FROM macro
- Okay, new message_from_mon function to hide some of the complexity in getting sessions, and checks for session != NUL...
- 11:50 AM Feature #10 (In Progress): osd: Replace ALLOW_MESSAGES_FROM macro
- Ugh, I thought Session was always set in the OSD but it turns out that's not the case.
- 11:50 AM Linux kernel client Bug #1 (Resolved): gpf in tcp_sendpage
- 11:50 AM Linux kernel client Bug #2 (Resolved): BUG at fs/ceph/caps.c:2178
- 11:44 AM Linux kernel client Bug #2 (Closed): BUG at fs/ceph/caps.c:2178
- 11:44 AM Linux kernel client Bug #2: BUG at fs/ceph/caps.c:2178
- Fixed by commit:e130642ba
- 11:50 AM Linux kernel client Bug #3 (Resolved): leaked dentry ref on umount
- 11:05 AM Linux kernel client Bug #3 (Closed): leaked dentry ref on umount
- 11:49 AM Linux kernel client Bug #22 (Resolved): BUG at fs/ceph/caps.c:253
- 11:04 AM Linux kernel client Bug #22 (Closed): BUG at fs/ceph/caps.c:253
- fixed by commit:d29b86892.
- 11:08 AM CephFS Bug #29: mds: rfiles underflow
- single mds.
- 11:08 AM CephFS Bug #29 (Resolved): mds: rfiles underflow
- [ERR] rfiles underflow -379 on [inode 10002154843 [...2,head] ~mds0/stray/10002154843/ auth v210197041 pv210219007 sn...
- 11:03 AM Linux kernel client Feature #27: ACLs
- Add ACL support in kclient. I suspect this just means wiring things up to the generic acl helper code (the actual ac...
- 09:48 AM Linux kernel client Feature #27 (Resolved): ACLs
- 09:49 AM Linux kernel client Bug #28 (Won't Fix): gracefully fail on fill_trace errors
- If fill_trace runs into problems when processing a reply, we should be sure to fail gracefully. Namely, if we can't ...
- 09:47 AM Linux kernel client Feature #26 (Rejected): statlite
- see -fsdevel thread from spring 09 for discussion (during LSF'09).
- 09:46 AM Linux kernel client Feature #25 (New): mdsc: mempool for cap writeback?
- 09:46 AM Linux kernel client Feature #24 (New): mdsc: preallocate reply msgs
- We should preallocate space for replies to our MDS messages.
- 09:45 AM Linux kernel client Feature #23 (Resolved): fcntl/flock advisory lock support
04/12/2010
- 10:31 PM Revision e7547dc7 (ceph): osd: add some put()s of Session * that weren't being used properly.
- This fixes a memory leak that could have caused problems in the future.
- 10:30 PM Revision 9ceb3be9 (ceph): osd: replace the ALLOW_MESSAGES_FROM macro with use of OSDCaps functions
- 10:17 PM Revision 4e138c1f (ceph): osd: Add is_X functions to OSDCaps (where X is mon, osd, mds).
- 06:20 PM Revision 17d24513 (ceph): rados: check that given pool name actually exists when deleting
- 06:20 PM Revision d44c2625 (ceph): mon: OSDMonitor makes sure pool ID is good, preventing crashiness
- 03:38 PM Feature #10 (Resolved): osd: Replace ALLOW_MESSAGES_FROM macro
- 03:36 PM Feature #10: osd: Replace ALLOW_MESSAGES_FROM macro
- Added is_mon(), is_osd(), is_mds() functions to OSDCaps. These use a new peer_type variable which is set by the OSD w...
- 02:25 PM Linux kernel client Bug #3 (Resolved): leaked dentry ref on umount
- fixed ref leak in dcache readdir, commit:2844a76a25
- 01:46 PM Bug #5 (Closed): ./rados lspools sometimes hangs after listing all pools?
- 10:51 AM Bug #5 (Resolved): ./rados lspools sometimes hangs after listing all pools?
- 01:46 PM Linux kernel client Bug #1 (Closed): gpf in tcp_sendpage
- 01:45 PM Linux kernel client Bug #1 (Resolved): gpf in tcp_sendpage
- 10:11 AM Linux kernel client Bug #1: gpf in tcp_sendpage
- Most likely kernel/module mismatch. Objdump didn't match oops message, I think this should be resolved-invalid unless...
- 01:07 PM Linux kernel client Bug #22 (Closed): BUG at fs/ceph/caps.c:253
- #12 0x0000000070c80b96 in ceph_add_cap (inode=0x6a5589b8, session=0x6fba8000, cap_id=236951, fmode=-1, issued=3413, w...
- 10:57 AM Feature #21 (Resolved): optionally use libatomic for atomic_t
- libatomic-ops on debian? something like that. unless there's something better.
also, probably drop the CommonC++... - 10:16 AM rgw Support #8: Document differences from S3
- We don't support HTTP POST at the moment.
04/09/2010
- 10:44 PM Revision c4892bed (ceph): osdmap: move pool names into first part of encoding; add version to sec...
- This gives the kclient access to pool names while still ignoring the
osd_info_t crap. And also will make adding more... - 10:13 PM Revision 6bf54f10 (ceph): librados: add lookup_pool()
- 09:52 PM Revision 6417e6a5 (ceph): librados: no need to include linux/types.h
- 09:52 PM Revision 26325b1a (ceph): debian: more librados1-dev deps
- 09:52 PM Revision 414a6e3e (ceph): spinlock: clean up #include deps
- 09:52 PM Revision 6ed1f04f (ceph): todo: move bugs to tracker
- 09:52 PM Revision 33e604dd (ceph): monclient: seed RNG so we pick a random monitor
- 07:35 PM Revision d5b39247 (ceph): mon: add 'get' command to AuthMon, letting you retrive one entity's key...
- 07:35 PM Revision 94307c9a (ceph): auth: cauthtool now identifies itself properly to common_init
- 06:53 PM Revision a9ecb46a (ceph): objectstore: fix clone_range() transaction encoding
- 06:50 PM Revision fa9588ab (ceph): filestore: fix weird references for clone/clone_range
- Not sure why that was legal to begin with?
- 05:04 PM Revision 1cfb41f7 (ceph): osd: rename PGPool::removed_snaps to cached_removed_snaps to clearly id...
- 05:02 PM Revision e949931e (ceph): mon: add 'get' command to AuthMon, letting you retrive one entity's key...
- 05:02 PM Revision e2ca9a30 (ceph): auth: cauthtool now identifies itself properly to common_init
- 04:26 PM CephFS Feature #20 (Resolved): client: recover from a killed session (w/ blacklist)
- The client_reconnect_stale config option no longer works because the blacklist/eviction logic changed significantly s...
- 04:15 PM Bug #5: ./rados lspools sometimes hangs after listing all pools?
- this was likely a msgr problem, fixed by commit:3d4a692b5.
- 12:35 PM Bug #5 (Closed): ./rados lspools sometimes hangs after listing all pools?
- This is rare and intermittent, but happens occasionally.
Best guess so far is that this locks the OSDMap but doesn't... - 04:12 PM Linux kernel client Feature #19 (Resolved): rbd
- blank
- 04:11 PM Linux kernel client Feature #18 (Resolved): reconnect fixups
- 02:02 PM Linux kernel client Bug #1: gpf in tcp_sendpage
- kernel dump didn't match running kernel.
saw this pop up immediately after starting bonnie.sh... but not after reb... - 10:34 AM Linux kernel client Bug #1 (Closed): gpf in tcp_sendpage
- [540049.710050] general protection fault: 0000 [#1] PREEMPT SMP
[540049.711610] last sysfs file: /sys/kernel/uevent... - 01:58 PM Bug #17 (Rejected): rm -r failure
- a few times i've seen an rm -r failure (on kernel tree, after untar + compile). something with the readdir from dcac...
- 01:58 PM CephFS Bug #16 (Resolved): mds restart vs dbench
- dbench 1, restart mds (may take a few times), dbench will error out.
- 01:57 PM CephFS Bug #15 (Resolved): mds rejoin: invented dirfrags (MDCache.cc:3469)
- invented dirfrags have version=0; subsequent modification of dentries/inodes predirty a bad (small) version #.
- 01:56 PM Bug #14 (Duplicate): osd: pg split breaks if not all osds are up
- 01:07 PM CephFS Feature #13 (Resolved): uclient: Make readdir use the cache
- Right now, readdir doesn't actually read out of the cache but sends a request to the MDS on every activation. This is...
- 01:04 PM CephFS Feature #12 (Resolved): uclient: Make cap handling smarter
- Right now the uclient isn't very intelligent about dropping capabilities. Change the cap drops associated with unlink...
- 12:56 PM CephFS Cleanup #11 (Resolved): mds: replace ALLOW_MESSAGES_FROM macro
- This is in MDS::_dispatch. This macro is used as a basic permission checker, but any permissions should be dealt with...
- 12:55 PM Feature #10 (Resolved): osd: Replace ALLOW_MESSAGES_FROM macro
- This is in OSD::_dispatch. This macro is used as a basic permission checker, but any permissions should be dealt with...
- 12:50 PM rgw Feature #9 (Won't Fix): Access unimported data
- Right now, rgw can only access data which has been added through rgw, even if a user's auid is set in rgw and matches...
- 12:46 PM rgw Support #8 (Closed): Document differences from S3
- RGW is a pretty good implementation, but it does have some differences from the S3 api. Document these.
They inclu... - 12:38 PM Documentation #7 (Resolved): Document Monitor Commands
- There's no centralized area to look for what monitor commands are available, or what permissions they require. (Of co...
- 12:36 PM Feature #6 (Rejected): libceph could use a backward-compatible-to function
- If this library is revised it'd be nice to have a way to check compatibility between versions.
- 10:58 AM Linux kernel client Bug #4 (Closed): lockdep warning in socket code
- [ 1615.328733] =======================================================
[ 1615.331050] [ INFO: possible circular lock... - 10:57 AM Linux kernel client Bug #3 (Closed): leaked dentry ref on umount
- After running misc qa stuff (bonnie.sh?), and umount, we get a crashed due to a leaked dentry reference.
ceph2 log... - 10:34 AM Linux kernel client Bug #2 (Closed): BUG at fs/ceph/caps.c:2178
- [563217.493362] ------------[ cut here ]------------
[563217.496350] kernel BUG at fs/ceph/caps.c:2178!
[563217.496... - 04:49 AM Revision b4c564c7 (ceph): librados: clean up object listing api
- 04:21 AM Revision bb2c0a92 (ceph): osd: clear initial interval_set in build_removed_snaps
- Caller may pass a non-empty set, in which case we crashed or returned a bad
result. - 03:55 AM Revision 1bcdf348 (ceph): testrados: create/use own pool
04/08/2010
- 11:42 PM Revision 0f22be7d (ceph): crc32c: ifdef out dead code to shut up warning
- 11:42 PM Revision b3cb1cb0 (ceph): Merge branch 'ns' into unstable
- 11:42 PM Revision c511dd1a (ceph): init-ceph: include 'noatime' btrfs mount option by default
- 11:42 PM Revision 736a2718 (ceph): filestore: more informative error on current/ rename failure
- 10:52 PM Revision 4a2ca2b1 (ceph): testrados: test more stuff
- 10:52 PM Revision c2987ce4 (ceph): objecter: cleanup
- 10:52 PM Revision f7f38df3 (ceph): librados: clean up c header
- 10:23 PM Revision 2112d208 (ceph): debian: fix up librados1-dev with all necessary headers
- 10:22 PM Revision 13d6b38c (ceph): librados: move c++ api into librados.hpp; more type cleanup
- 09:19 PM Revision 1e05d5c0 (ceph): librados: use std::string, not nstring, in c++ api
- 08:00 PM Revision 181b81c1 (ceph): librados: avoid snapid_t type, types.h
- 07:57 PM Revision 6c9b36b2 (ceph): rbdtool: fix man warnings
- 05:39 PM Revision 5c00d5f7 (ceph): filestore: check for SNAP_DESTROY ioctl, and file if not present
- 04:54 PM Revision c9cb1183 (ceph): librados: move librados.h to include/
- 04:51 PM Revision 2fae333d (ceph): spinlock: move to include/
- 04:49 PM Revision ee13f60f (ceph): spinlock: move to ceph namespace
- 04:48 PM Revision e1896d8d (ceph): atomic: move to ceph namespace
- 04:40 PM Revision 3ea82baa (ceph): atomic: rename test() -> read()
- Make the interface slightly less weird.
- 04:16 PM Revision aa55570d (ceph): buffer: avoid including ceph assert directly
- If assert isn't defined, include the standard C one. Internal ceph bits
should include assert.h before buffer.h if t... - 04:09 PM Revision df2422f3 (ceph): assert: use ceph namespace
- 04:04 PM Revision 74be6771 (ceph): crc32c: prefix with ceph_
- 03:53 PM Revision ac70c563 (ceph): page: move to ceph namespace, defs to common/page.cc
- 03:49 PM Revision fc8e3c35 (ceph): buffer: move to ceph namespace
- 03:13 PM Revision 50ea193c (ceph): osd: fix transposed clone_range args
- We're cloning ranges _into_ soid (and in fact just deleted any old soid
that may have been lingering). - 02:43 PM Revision 250ff72a (ceph): spinlock: avoid lockdep include unless lockdep is on
- 02:43 PM Revision 0fb8d931 (ceph): buffer: avoid spinlock include unless debugging is on
- 02:42 PM Revision 75699dbb (ceph): librados: reduce #includes
04/07/2010
- 11:43 PM Revision 7f16acca (ceph): librados: c binding for rados_tmap_update()
- 11:37 PM Revision e5381a5c (ceph): rbdtool: maintain an rbd_directory object; add --list
- 11:35 PM Revision a1cf10ab (ceph): auth: shut up about loading the key file
- 11:35 PM Revision be1e87f3 (ceph): librados: add tmap_update to c++ interface
- 11:35 PM Revision ba584126 (ceph): osd: fix tmap when header update not specified, or initial object dne
- 10:29 PM Revision 6e28eb3d (ceph): Merge branch 'rbd' into unstable
- 10:29 PM Revision b89e743c (ceph): debian: include rbdtool
- 10:29 PM Revision 3d7848da (ceph): rbdtool: add man page
- 10:29 PM Revision 25760897 (ceph): rbdtool: delete rbd images too
- 09:47 PM Revision 89b05587 (ceph): rbdtool: use shared rbd_types.h
- 09:23 PM Revision 3d4a692b (ceph): msgr: fail on sd < 0
- I saw a case where poll(2) was blocking despite being passed an fd of -1.
Since that's clearly invalid, we can fail t... - 09:23 PM Revision d2d80c6e (ceph): ceph: print output to stdout, not dout
- Using cout instead of dout(0) avoids the thread id in the prefix.
- 09:23 PM Revision 50fcf793 (ceph): msgr: drop useless was_session_reset() call on lossy channel close
- The fail closes out the pipe, no need to do a 'reset'.
- 07:38 PM Revision 686192a0 (ceph): osd: flush store ops after map update
- Otherwise we can race:
- queue delete+rewrite of pg log
- unblock op queue
- queue append to log
...
- apply app... - 05:52 PM Revision 2f210106 (ceph): mds: set laggy in tick() if its been a while
04/06/2010
- 11:18 PM Revision 56de6622 (ceph): debian: drop empty dir from radosgw
- 09:54 PM Revision 8741959b (ceph): todo: mds
- 08:30 PM Revision 03e20194 (ceph): mon: set osd up_thru when adjusting pg_temp
- This just avoids making the OSD ask again for up_thru. Fewer osdmap
updates! - 07:58 PM Revision 634441e4 (ceph): osd: distinguish between per-pool snap pools and user-managed snap pools
- If removed_snaps is non-empty, then snaps are managed by the
user: snap context is specified for all writes (e.g., MD... - 07:32 PM Revision 2c2ea0a2 (ceph): todo: updates
- 07:32 PM Revision 07e39846 (ceph): mds: journal oldest client tid
- Journal the client's safe tid with new requests. This keeps the client
completed_requests list trimmed, so that we d... - 07:32 PM Revision 4de289f5 (ceph): mds: use send_dentry_unlink helper for remote unlink
- 07:32 PM Revision 985a9810 (ceph): mds: move new stray dentries to end of lru
- 07:32 PM Revision 50b17728 (ceph): osd: return more useful errno (ENOENT) on missing base
- 07:32 PM Revision 0efe4fbd (ceph): mds: fix ref count check in eval_stray, purge_stray_purged
- Dentry gets pin for any inode pins, and on purge has PIN_PURGING.
- 07:32 PM Revision be4a797e (ceph): mds: fix purge_stray to be smart about NEW dentries
- Mark new stray dentries NEW.
After purge, remove new dentries immediately; leave others as NULL and
dirty so they ge... - 07:32 PM Revision e52aa5e8 (ceph): mds: remove null clean stray dentries after dir commit
Also available in: Atom